Trawl docs
Trawl.
AI web extraction — public pages in, schema-validated JSON out.
Trawl runs asynchronous extraction jobs: you describe what you want and the shape you want it in (a JSON Schema), and an AI agent driving a real browser finds it on the public web. The result is validated against your schema before it's delivered — poll the job, or receive a signed webhook. This documentation covers everything from your first API call to production operations.
Browse
What you’ll find here
Quickstart
Go from zero to your first extraction in minutes. Create a job with a JSON Schema and a description, poll it, and read the structured result.
Read more Core concepts
Understand jobs, the extraction agent, result schemas, the job lifecycle, and how webhooks deliver results to your endpoint.
Read more API reference
Complete endpoint documentation with request and response examples for jobs and webhooks in the Trawl API.
Read more Guides
Step-by-step walkthroughs: foundations, your first extraction, schema design, signed webhooks, and running Trawl in production.
Read more Reference
Base URL
https://api.trawl.productcraft.co/v1All endpoints are versioned under /v1 and scoped to a workspace: /v1/workspaces/:workspace_id/.... Request and response JSON uses snake_case keys throughout.
Reference
Authentication
Authenticate with a workspace API key, minted at console.productcraft.co, sent as a bearer token:
Authorization: Bearer pcft_live_...- Workspace-scoped access. Every route lives under your workspace, and routes are gated by
trawl.*permission actions. Your workspace must have thetrawlservice enabled. - Typed SDK or plain HTTP.
@productcraft/trawlon npm is a typed Node/TypeScript client generated from this API's OpenAPI spec. The docs show curl and TypeScriptfetchexamples throughout — every endpoint is reachable either way.